Α comprehensive geotechnical characterisation of overburden material from lignite mine excavations

For the sustainability of lignite and coal mining, engineers need the soils' appropriate geotechnical characterisation of the excavations. This characterisation relates to the slope stability, the overall area stability, and the sustainable exploitation of mining areas during operational and post-closure eras. This work presents a novel perspective on the soil characterisation of lignite mine excavations. Rather than focusing on one particular mine, a comprehensive geotechnical characterisation provides insights about the soils of related mining areas. Additionally, slope stability representative cases guide soil characterisation and analysis rather than following it. The Greek lignite mines are used as a representative example for the illustration of this framework. Initially, the two typical slope stability scenarios for Greek lignite mines are established: a slope in the presence of a weak zone and a homogeneous slope. The crucial geotechnical parameters are then statistically examined based on an extensive database established from laboratory results of various mines. Focus is on shear strength (friction angle and cohesion), the key to slope stability, and physical properties that can provide the strength through empirical correlations. Many soil layers exhibit a low residual friction angle from 5° to 15°, corresponding to the weak zone. Peak friction angle presents a mean value of 28.4°, and a characteristic value of 25.0°; cohesion presents a wide range with a mean value of 111 kPa and a characteristic value of 84 kPa. Finally, cross-correlations between geotechnical parameters present large scatter, but relations for estimating the residual and the peak friction angle reveal general trends.


Introduction
The exploitation of coal and lignite (brown coal) has defined the global energy mix for decades. Coal is a geo-resource that has contributed significantly to the global energy production accounting for 24.5% of the global energy production in 2000, 30% in 2010and 27% in 2019(BP 2020Ritchie 2014). Several countries (e.g. Canada, Germany, France, Italy, the United Kingdom, and Greece) have pledged to phase out all coal-powered electricity production in the following decades. After the mines are closed, stakeholders and beneficiaries should reclaim these vast mined lands with solutions ranging from simple vegetation and flooding of pits to photovoltaic and wind parks (Kasztelewicz 2014;Szczepiński et al. 2010). Coal and lignite mining can remain commercially profitable for the few remaining years, only if mining operations are optimised and efficiently managed.
When coal and lignite seams are located relatively near the surface, they are usually extracted by open-pit mining methods (also referred to as opencast or opencut). During the last decades, large areas have been excavated, creating deep slopes (e.g. up to 200 m for Greek, Australian or Indian mines), and vast amounts of overburden materials have been dumped within or outside the pits (e.g. Leonardos and Terezopoulos 2003;Scott et al. 2010;Verma et al. 2013). Geotechnical characterisation and stability of coal and lignite surface mines have been and will continue to be of significance. Issues related to mines' stability will also remain of interest during the post-lignite era, as the large mining areas will adapt to future uses.
One critical problem for the effective management of surface coal and lignite mines is slope stability. Many incidents of excessive displacements and slope failures have been reported in the literature (Kavvadas et al. 2013;Marinos et al. 2015;Satyanarayana and Sinha 2018;Tutluoglu et al. 2011;Zevgolis et al. 2019;Zhigang et al. 2020). In this context, slope stability is crucial for the personnel's safety, the efficient operation of the mines, and the area's stability. Moreover, slope stability dominates in mine reclamation, becoming the critical factor in decisionmaking processes (Kasmer et al. 2006;Poulsen et al. 2014;Sharma and Roy 2015;Steiakakis et al. 2009). Thus, an in-depth evaluation of the soil properties is necessary for a stability study due to the analysis's sensitivity on material characterisation (Ghadrdan et al. 2020).
The present work suggests a novel framework on the soil characterisation of lignite mine excavations. Rather than focusing on one particular mine (e.g. Bednarczyk 2017;Pulipati et al. 2020;Ulusay et al. 2014), a comprehensive geotechnical characterisation provides insights about different but related mining areas. Moreover, slope stability scenarios are employed to guide soil characterisation, contrary to typical analysis where characterisation always precedes slope stability. This approach differs from classical approaches without substituting them; combining and using existing data, which were obtained to analyse various mines, can provide an overview and additional knowledge of related mining areas leading to better design. Thus, this work's approach is of significance for stability and sustainable operation and reclamation of mining areas, and broader importance in the field of geo-resources and geo-energy.
To illustrate this framework, the geotechnical properties of Greek lignite mines are studied within a unified context, as a complementary tool to typical processes that focus on the particular geology of an area or a mine. The slope stability scenarios guide the characterisation towards the crucial parameters and their proper analysis.
Results from several laboratory investigations, referring to different mines but with similar expected slope stability mechanisms, are used to compile an extensive database; the ultimate purpose is to characterise the mines' soil in a unified way. Focus is on residual strength (residual friction angle) and peak shear strength (peak friction angle and cohesion), which are the critical parameters to the two slope stability cases that will be discussed. Physical properties are also thoroughly analysed as they are most commonly and easily measured, they characterise the soil types and they can provide the shear strength through empirical correlations. The inclusion of different mines increases the statistical significance of the results and provides an overview of soil's properties. Based on this framework, the soil materials' engineering properties can be practically evaluated more clearly. Finally, this overview can be employed as a guide for preliminary analysis; furthermore, it can improve the design methodologies and expand the knowledge on lignite excavations' slope stability.
At the following sections, initially, a brief background of the Greek lignite mines is presented. Moreover, based on previous research and experience, the major slope stability scenarios are discussed, applying to most Greek lignite mines. Furthermore, physical and engineering properties are statistically analysed; this statistical analysis can back feed the slope stability analysis of lignite mines and the stability and geotechnical analysis of the mining areas. Finally, cross-correlations between geotechnical parameters are evaluated, and relevant conclusions are discussed.

Greek lignite mines
Lignite has dominated in Greece, being a geo-resource in large quantities that has contributed considerably to the energy production and domestic electricity demand (EURACOAL 2020). Greece has traditionally been a lead lignite producer and has largely covered its energy needs via lignite. This energy source accounted for the country's supplies as much as 60% in 200460% in , 49% in 201160% in , and 29% in 201960% in (Eurostat 2020. Over the years, as the shallow deposits have been exploited, lignite mines become extremely deeper and the corresponding excavations extremely large. Exploitable lignite deposits are principally located in Western Macedonia, northern Greece, and secondarily in the Peloponnese, southern Greece. The major basins in the first case are Ptolemais-Amyntaio, Florina and Kozani-Servia, while in Peloponnese the lignite basin is in the Megalopoli area. In this work, laboratory results from samples originating from eleven different mines are employed. Ten out of eleven mines are located in Western Macedonia and its three major basins; Amyntaio, Field 6, Kardia, Komnina, Main Field, Mavropigi, South Field, Vevi in the Ptolemais-Amyntaio minefield, and Prosilio, Servia in the Kozani-Servia minefield. One mining complex (Megalopoli) is located in Peloponnese.
The typical ground profile includes a surface zone of sterile materials (several tens of meters on average) and a deeper zone, where the exploitable lignite deposits lie. Thin layers of sterile materials can be commonly found within these deposits (e.g. Monopolis et al. 1999). The sterile materials are soil-rock deposits typically consisting of marls and stiff or hard clays, while weaker cohesive soils and sands can also be encountered (e.g. Kavvadas et al. 2020;Koukouzas 2007). Therefore, soil mechanics principles and soil characterisation (rather than rock mechanics and rock mass characterisation) are relevant for slope stability issues.
It is noteworthy that marls' engineering behaviour and response require special attention, being crucially affected by water content changes. Thus, typical geotechnical evaluation of stiffness and strength should be applied with caution.
Moreover, on some occasions, mild tectonics of mining areas caused a constant orientation of horizontal or sub-horizontal layers that might dip with a small angle (0°-6°). Specific tectonics might lead to complex failure mechanisms, which are not considered in this work because the effect of an active fault to a coal mine is case-specific (e.g. Driad-Lebeau et al. 2005). Many researchers have assessed the geologic formations (Koukouzas 2007;Pavlides 1985) and ground profiles of Greek lignite mines from a geotechnical perspective (Kavvadas et al. 2020;Zevgolis et al. 2019), and thus, a more in-depth analysis is not presented herein.
Over the last three decades, surface mining was practically the only practice in Greek lignite mines. Typically, during the open-pit lignite mining, the thick upper zone of sterile materials is excavated, and the main body of the lignite is exploited. The continuous mining method is employed: large bucket-wheel excavators operate on the slopes on benches, and a system of conveyor belts transports the lignite and the sterile materials. These are either transported and dumped by spreaders at nearby areas forming massive spoil heaps or backfill mine areas on which operations have ceased. Figure 1 shows an indicative and simplified cross-section of a deep excavation.
Although continuous mining is the norm, conventional mining methods (hydraulic excavators, loaders, and dumpers) are also employed when required, on a smaller scale. These methods are even more common on smaller, private mines, e.g. Lava and the Prosilio mines (Kozani-Servia basin). Finally, controlled use of explosives (drill and blast) is employed when very hard layers are encountered (e.g. in the South Field mine, in the Ptolemais-Amyntaio minefield).
Due to the open surface character of the Greek lignite mines, slope stability is a critical issue. Several past failures demonstrate that slope stability has not always been rigorously evaluated, causing problems in safety and management of the mines (Zevgolis et al. 2019). As a sequel to Zevgolis et al. (2019), an overview of the geotechnical characterisation of the soil materials found in Greek lignite mines is herein presented.

Overview
In this work, a comprehensive approach is followed to identify soil materials of Greek lignite mines. A particular mine, geometry or geology is not identified, but general patterns and characteristics are extracted. An extensive database was established based on laboratory experiments on soil samples of various mines in Greece (in the regions of Western Macedonia and Peloponnese). Few of the database results have been published by other researchers (Leonardos 2004;Steiakakis 2003) and are included in this work for completeness, but the vast majority of data are provided for the first time.
Initially, soil's strength in terms of residual friction angle is examined with depth. Residual strength, i.e. the shear strength at very large deformations (Holtz et al. 2010), is chosen because it characterises lignite mines' slope failure and typical laboratory results in Greek mines include this type of strength. Two major scenarios are distinguished, as depicted by two indicative boreholes in Fig. 2. Figure 2a  From these two scenarios, the critical case for slope stability is when a zone of low strength appears. This zone, named herein as weak zone, could be a ''weak'' layer or an interface between marls, clays, or lignite at different combinations (e.g. marl to marl, marl to clay). This zone is usually thin and varies from several centimetres to a few meters. Moreover, it has an unloading elastic modulus much lower than the adjacent layers. As a result, large shear strains might develop during deformation and soil usually fails on its residual strength. More details on this zone's geological origins and its failure state were provided by Kavvadas et al. (2020). The overburden soil becomes of secondary importance, and identifying the residual strength of the weak zone is critical. This has also been confirmed by back analyses of documented slope failures (Kavvadas et al. 2013;Prountzopoulos et al. 2017;Steiakakis et al. 2017;Zevgolis et al. 2019). Typically, residual strength is measured by the ring shear test, and if presented with depth, it can reveal the profile of the weak zone (as in Fig. 2a).
A second scenario refers to the case where a distinct weak zone is not present. This scenario is usually less sensitive in terms of stability, and for engineering routine purposes it considers a practically homogeneous slope overlying a deep bedrock formation (in this case, identifying and separating the geological layers is insignificant (Fig. 2b). This scenario is less conservative but should also be examined, because under certain circumstances the slopes can fail, e.g. if pore pressures suddenly increase. Figure 3 demonstrates two cases of slope stability failures whose mechanisms were discussed by Leonardos (2004) and represent the two case scenarios' failure surfaces. It is underlined that the complete stratigraphy of the slopes of lignite mines is more  (2004)) complicated, as discussed in Sect. 3 than the simplified description provided herein. However, only the overview of the geotechnical properties of the different layers crucially affects slope stability.

Classification and index properties
Many samples were gathered from 11 Greek lignite mines (Amyntaio, Field 6, Kardia, Komnina, Main Field, Mavropigi, Megalopoli, Prosilio, Servia, South Field, and Vevi). The statistics of Atterberg limits, soil wet unit weights, and water contents are summarised in Table 1. Furthermore, Fig. 4 presents a few characteristic granulometries of fine-grained materials (dominating all mines) from different depths (1.5 m to 100 m). Coefficient of uniformity (Cu = d 60 /d 10 ) is an indicator of the shape of the distribution; soils ranging from poor-graded to well-graded are present with Cu varying from 4.2 to 18.9. ASTM C136, ASTM D4318-83, ASTM D854-14 and ASTM D 2216 standards were followed.
According to the USCS most soil samples are characterised as fine-grained, i.e. silts and clays (in the database, 1117 samples out of 1163 are fine-grained and the rest are organics and silty or clayey sands), agreeing with the general geological identification (see Sect. 3). Figure 5 presents the distribution of the samples based on USCS classification. The 118 samples not classified as silts or clays are mainly organics, and few are silty and clayey sands.
In order for the fine-grained spoil material to be classified, Liquid Limit (LL) with Plasticity Index (PI) were plotted for all samples (Fig. 6). Again, clays and silts of high and low plasticity are present with silts (mostly marls) being most frequent. However, based on different mines, there are cases where clays are dominant, or where silts are dominant and others where both clays and silts are similarly encountered. Figure 7 presents the histograms of LL and PI. The Liquid Limit (Fig. 7a) has an approximately normal distribution, with a peak at 50-55% and a mean value of 54%; this can be additionally assessed in Fig. 6. Furthermore, Plasticity Index (Fig. 7b) presents results close to a truncated normal distribution with values gathered around the mean value of 18% and a long right tail (with values up to 70%). Plastic Limit (PL) (Fig. 7c) has a distribution that fits well a lognormal one with a mean value of 35%. Water content (w) (Fig. 7d) is well fitted with a normal truncated distribution and a mean value of 40%, close to the mean PL. The COVs of LL, PL, PI, and w are 26%, 37%, 57%, and 39%, respectively. As expected, PI's COV is the largest, given that it is a function of LL and PI (so the uncertainty is larger than the largest of the two). In any case, these values are not uncommon for natural soils, as fine-grained materials may often have COVs of 30% or more for LL and PL, and up to 80% for PI (Baecher and Christian 2003;Lacasse and Nadim 1996).
For 649 out of 1128 samples (57% of the sample population), LL was larger than 50 (i.e. samples of high plasticity). On the other hand, in terms of water content, w lied between PL and LL in only 512 out of 819 samples (total number of samples with simultaneous PL, LL and w measurements). That signifies that 62.5% of the samples lie somewhere between the typical plastic solid-state range (i.e. PL \ w \ LL). Moreover, w [ LL for 94 samples and w \ PL for 213 samples; in other words, a noteworthy part of the samples (about 11%) can be considered as a soft normally consolidated material, while another Another standard index used in scaling the natural water content of soils is the Liquidity Index, defined as LI = [w -PL]/[LL -PL]. Typical cohesive soils have LI values within the range 0 to 1 (i.e. PL \ w \ LL), which means that a soil behaves plastically, or in a mouldable way. In principle, LI approaching 0 or 1 indicates that the soil is in a very stiff or very soft state, respectively (Murthy 2002). Based on the available data, LI's mean and median values were computed equal to 0.41 and 0.25, respectively. These values declare a, per average, slightly overconsolidated soil. In light of the above, Fig. 8 shows a water content continuum and the relevant soil states, together with indications of consolidation degree and generalised stress-strain responses. By examining the whole population, LI ranges between 0 and 1 (typical behaviour) in 501 samples (61%); as already stated, this is a typical range indicating soils that behave plastically. In 218 samples (27%), LI is smaller than 0; this is typical for over-consolidated to heavily overconsolidated material (Fig. 8e), and from a stressstrain perspective, it means that the soil's behaviour will be brittle if sheared (Fig. 8d). Last, in 100 samples (12%) LI is larger than 1; this indicates a normally consolidated, or even sensitive cohesive soil (Fig. 8e). If sheared, this type of soil behaves as a very viscous liquid (Fig. 8d). Holtz et al. (2010) (Smith 2014;Terzaghi et al. 1996). It is noteworthy that LI presents a vast range and an extremely large COV; this happens mainly due to the presence of semi-solid and solid specimens with w significantly smaller than PL, and ''sensitive'' samples with LI [ 1. Moreover, LI depends on LL, PL, and water content; thus, error in these values propagates in LI, leading to a high COV. Few literature data can be found regarding LI's COV, one reason being that for low mean LI values (approaching zero), COV becomes very large even with a small standard deviation.
Basic statistics for moist and dry unit weight are also provided in Table 1. The mean values of the moist and dry unit weight are relatively low (17.2 kN/m 3 and 12.1 kN/m 3 , respectively), indicating the soils' finegrained character. It has been reported that regardless of the type of soil, unit weight, and density demonstrate a small variability compared to most other soil properties, typically less than 10% (Baecher and Christian 2003; Lacasse and Nadim 1996). The COV of unit weights in this work is 9% and 14% for moist and dry unit weight, respectively. Void ratio presents a mean value of 1.20 and a relatively high COV of 38%, while Lacasse and Nadim (1996) suggest COV values for void ratio up to 30%.
Finally, a very strong negative cross-correlation is noted between initial void ratio e and moist unit weight c, corresponding to a linear correlation coefficient of q = -0.88 and a determination coefficient R 2 = 0.77 (Fig. 9). It is well known that e and c are dependent, also through water content w and the solid grains' density, which usually does not vary significantly. Although the water content has a noticeable variation (COV w = 39%) a correlation between e and c exists, and the one can be directly calculated from the other as c = -4.2e ? 22.4. This is a strong linear correlation based on the dependency of the two parameters and the large R 2 value.

Engineering properties
Shear strength is the critical property related to slope stability. Based on the two case scenarios previously described, two sets of shear strength parameters are important: those characterising the peak shear strength and those defining the residual shear strength. These parameters have been investigated with triaxial and direct shear tests for the peak strength (ASTM D2850 and ASTM D3080), and with ring shear tests for the

Residual strength
Residual strength is characterised only by the friction angle U r , as the cohesion is essentially zero at this extreme state of large strains. Figure 10 presents the histogram of residual friction angles from 192 samples and depicts two peaks, approximately 10°and 26°. Notice that the test samples came from several depths and were not targeted on the weak zones. Thus, these residual angles describe not just the weak zone, but all soil formations. In that vein, two areas are distinguished in the histogram of Fig. 10 separated at 15°; this limit separates two areas with a similar number of samples and is considered a reasonable upper value for the weak zone's residual strength. Moreover, the statistics of the two areas of the histogram (weak and non-weak zones) are introduced in Table 2. The mean value for the weak zone is 10°with a standard deviation 2.7°. The weak zone's friction angle can vary from 4.6°to   Fig. 7 Histogram of a Liquid Limit (LL), b Plastic Limit (PL), c Plasticity Index (PI), and d water content (w) of fine-grained overburden soil materials 15°, i.e. about two standard deviations around the mean value. Finally, Table 3 displays the mean values for each mine's residual friction angle (for all soils, and the two zones identified above). Some mines exhibit mean values of residual angles slightly larger than the total mean, while others slightly smaller than the total mean; however, there is no mine whose values significantly diverge from the total mean.  In agreement with previous literature (e.g. Lambe and Whitman 1969;Terzaghi et al. 1996) and engineering practice, an effort was made to evaluate the relationship between the residual strength and the physical parameters, which are more easily and frequently measured in practice. The large scatter of results does not allow for an accurate correlation; nonetheless, the residual friction angle can be roughly estimated for a preliminary evaluation based on the Plasticity Index. Figure 11a illustrates this correlation and an ''average'' relation U r = 265PI -0.9 is proposed, along with an upper and a lower limit. The significant scatter limits the applicability of this relation to a preliminary analysis, and the two boundaries should always be considered alongside it. For this analysis, the soils have been broadly distinguished based on their USCS classification (clays, silts, or organics), but this does not decrease the scatter or propose a different trend.
Finally, Fig. 11b illustrates residual friction angles with Liquidity Index. No clear relation can be identified, but most LIs lie between zero and 0.5 in agreement with statistical results in Table 1.

Peak strength
Peak shear strength crucially affects stability. In the presence of a weak zone, peak strength (rather than residual) would be the overburden soil's actual strength, i.e. all the soil formations except the weak zone. In a homogeneous slope scenario, the soil's engineering response is simulated based only on its peak strength, not on its residual one. In this work, the soils' peak strength has been quantified by triaxial and secondarily by direct shear tests. Triaxial is an advanced experimental device that can quantify the stress-strain response of a soil sample, while the direct shear test is sometimes considered more appropriate n population, l mean value, m median value, r standard deviation, COV coefficient of variation Table 3 Mean value for residual friction angle (r' v &700 kPa) for all materials, for weak (B 15°) and non-weak ([ 15°) zones, peak friction angle and cohesion for different mines based on triaxial tests for slope stability analysis. Twenty-two (22) triaxial consolidated drained, 130 consolidated undrained (with pore pressure measurement) and tests 52 direct shear tests are used. The Servia mine presents noticeable differences compared to the other mines, as will also be revealed in the following. Figures 12  and 13 present friction angle and cohesion obtained from triaxial and direct shear tests; the results are distinguished between the Servia and the other ten mines.
Initially, the focus is on the ten mines (excluding Servia); the distribution of the friction angle for triaxial tests approaches a normal distribution with a mean value of 28.4°, ranging from 14°to 44°, and a COV equal to 24%. Although this range appears large, most results (77 out of 112 samples, i.e. about 70%) lie within one standard deviation around the mean value. In current design frameworks (e.g. Eurocode 7 (EC7) or Load and Resistance Factor Design (LRFD)), the mean value is not directly used for analysis and design.  Fig. 12 Histogram of a friction angle and b cohesion from triaxial tests (statistics mentioned refer to the ten mines without Servia). n: population, l: mean value, m: median value, r: standard deviation, k: characteristic value, COV: coefficient of variation For example, the EC7 framework proposes the socalled characteristic value X k of a soil property. Schneider (1997) demonstrated that an appropriate approximation of X k is one half a standard deviation below the mean (X k = l -0.5r); this definition has gained wide acceptance within the geotechnical community (Orr and Farrell 2000). So, based on this definition, the characteristic value of the friction angle is 25.0°. On the other hand, cohesion (excluding Servia) presents a vast range, from 0 to 631 kPa, with a mean value 185 kPa, a median value 145 kPa and a COV 80%. This range is due to the cohesion's nature, and the wide range of materials and stress ranges used. Based on Schneider's definition, the characteristic value c' k drops to 111 kPa, considering the many samples with low or zero cohesion.
The Servia mine presents noticeable differences from the other mines. Only for this mine, the mean value for the peak friction angle is 41°(compared to 28.4°of the other ten mines), and for the cohesion is 242 kPa (compared to 185 kPa of the other mines). If the peak friction angle is evaluated based on each mine (Table 3), it becomes even more apparent that this mine has a significantly larger value than the rest. Finally, notice that the Servia mines' residual strength was unavailable and not used in Sect. 4.3.1 and results from the Servia mine have a minor influence on classification and index limits.
Few direct shear tests were obtained and only for two mines, Mavropigi and Servia; these show a different distribution of the friction angle ( Fig. 13a) but cannot be straightforwardly evaluated due to their small population. The Mavropigi mine presents a mean value of 24.7°and a standard deviation of 5.3°a nd the Servia mine a mean value of 37.7°and a standard deviation of 12.3°. This comparison further validates the different nature of the soils of the Servia mine. A smaller mean cohesion is presented for direct shear than for triaxial tests (72 kPa vs 200 kPa for Mavropigi and 131 kPa vs 242 kPa for Servia).
Empirical correlations between peak friction angle and physical properties of soils are common and were investigated in the present study. Figure 14a illustrates such a correlation between peak friction angle and Plasticity Index. Notice that almost all silts with PI \ 12% come from the Servia mine but do not change the overall trend. Substantial scatter is observed, as in any correlation between peak friction angle and physical parameters of this work. As a result, well-known relations (e.gKulhawy and Mayne 1990; Lambe and Whitman 1969;Terzaghi et al. 1996) do not fit the observed data well.
However, an average relation / 0 = -20log(PI) ? 56 is proposed but should be applied with caution (Fig. 14a). This correlation presents a large scatter and could be employed only in a preliminary analysis with respect to the presented boundaries. Figure 14b shows   13 Histogram of a friction angle and b cohesion from direct shear tests. n: population, l: mean value, m: median value, r: standard deviation, k: characteristic value, COV: coefficient of variation the peak friction angle with Liquidity Index, with the samples broadly categorised as clays, silts, and organics (based on USCS), similarly as in Fig. 11b. A correlation cannot be established; all points with LI less than -0.5 come from the Servia mine, depicting its difference with all other mines. Essentially, all mines (except Servia) present LI values from -0.5 to 1 and a peak friction angle from 15°to 45°; Servia mine presents samples with LI values even less than -0.5 and friction angles from 15°to 60°. Finally, the authors evaluated several types of correlations between the geotechnical parameters (e.g. friction angle and cohesion, Atterberg limits, unit weight, water content, and void ratio). However, no strong correlations (except the one of Fig. 9) were established. Indicatively, the cross-correlation between peak friction angle and cohesion is evaluated in Fig. 15.
Natural soils' friction angle commonly decreases with the decrease of the cohesion, a trend also observed here. However, no cross-correlation is established as the scatter is exceedingly high. An upper limit is / 0 = 60-0.6c 0 , and all friction anglecohesion combinations lie below this line, having a linear correlation coefficient of q = -0.21. It shall be noted that the coefficient of linear correlation specifically for triaxial CD tests was -0.80 (22 samples), while for triaxial CUPP tests was -0.12 (130 samples) and for direct shear (DS) tests -0.23 (52 samples).

Conclusions
In this work, an overview of Greek lignite mines' geotechnical characterisation was presented from a slope stability perspective, presenting a novel analysis framework. An extensive database based on geotechnical investigations from 11 Greek lignite mines has been established, and the results were statistically treated and combined with previous knowledge and The results of such an analysis can be used in a preliminary design stage and provide an overview of the expected soils and their geotechnical properties. Besides, following this analysis, significant errors can be avoided in design and management. However, the particular geology (including tectonics), stratigraphy and soil properties of a specific mine, not offered by the general overview proposed herein, must always be considered in the management and design decisions. Initially, the physical properties of the examined samples were analysed according to geotechnical standards. Index and classification properties describe fine-grained materials (silts and clays), consistent with known geological formations of Greek lignite mines.
The mean values of the Liquid Limit and the Plasticity Index are 54% and 18%, respectively. The mean value of the Liquidity Index (LI) is close to zero and indicates a plastic soil close to its semi-solid state. Nonetheless, LI presents an exceptionally large range and a large COV.
Furthermore, two critical slope stability scenarios for lignite mines were employed to define the analysis framework: (i) a slope with a weak zone of low strength (ii) and a homogeneous slope. Residual strength is the essential parameter for the weak zone, and peak strength for the homogeneous slope. The soil strength was evaluated through ring shear tests for the residual strength (residual friction angle) and triaxial and direct shear tests for the peak strength (peak friction angle and cohesion).
Measurements show that many soil layers present a very low residual strength-friction angle from 4.6°to 15°-corresponding to the weak zone. In contrast, for all other cases, higher residual strength is measured. Furthermore, the peak friction angle from triaxial tests presents a mean value of 28.4°with most results lying within one standard deviation around the mean value. The characteristic value, as defined by Eurocode 7, becomes 25.0°. Cohesion based on triaxial tests presents a broad range, from 0 to 631 kPa with a mean value of 185 kPa, and the characteristic value is 84 kPa (reflecting the fact of numerous specimens with zero or low values). One mine (Servia) presents an overall different strength, significantly higher than the other mines. Thus, it was individually evaluated having a mean friction angle of 41°and a mean cohesion of 242 kPa.
Finally, several joint distributions between geotechnical parameters were evaluated. All relations between physical and engineering parameters present large scatter, and strong correlations are not observed. However, relations for a rough estimate of the residual and the peak friction angle were proposed and reveal general trends that can be used on a preliminary analysis with due caution.
This work aims to create a framework for stability analysis of lignite mines in the context of preliminary analysis and design and/or in the absence of sitespecific geotechnical investigation. Greek lignite mines' geotechnical characteristics have been evaluated based on an extensive laboratory test database to illustrate this framework. This characterisation relates to mines' slope stability, the overall stability, and the geotechnical analysis of mining areas during their current operational and post-closure eras.